IBM carbon nanotube transistors could make CPUs five to ten times faster than today's silicon chips
IBM have demonstrated a new way to place carbon nanotubes as transistors in the commercial production of the teeny, tiny and freakishly powerful, processors of the future.
Microchip manufacturers have been looking for ways to keep up with the demand for ever smaller, ever quicker, ever more efficient chips to power all our techie devices - and they have been very successful. But we are getting to a point where the limits put in place by the laws of physics are going to get in the way of further generations of our gaming processors.
The current smallest chip design in your desktop PC sits at 22nm with the next shrink down to 14nm in 2014 with Intel’s Broadwell CPUs. At the Intel Developer Forum this year, their head transistor-shrinking guru, Mark Bohr, explained they’d got a solution for shrinking down the silicon to 10nm and ideas for moving all the way down to 5nm.
But as you get smaller things like physics get in the way and using silicon to fashion your teeny transistors gets almost impossible. If Moore’s Law is to carry on something needs to change.
And carbon nanotubes offer something that can be shrunk way past the limits on silicon, and also offer much, much better performance. IBM’s modelling would suggest that a chip made with carbon nanotubes could offer five to ten times improvements in performance compared to silicon chips.
But until recently the advantages of using this material were outweighed by the fact it was only really manageable in laboratories, and even then scientists could only manage to place a couple hundred together at a time. With modern CPUs, such as the Ivy Bridge chips from Intel, hitting 1.4 billion transistors in a chip these scientists need to start thinking bigger. Well, smaller. But with more bits.
“Carbon nanotubes, borne out of chemistry, have largely been laboratory curiosities as far as microelectronic applications are concerned,” said Supratik Guha, Director of Physical Sciences at IBM Research. “We are attempting the first steps towards a technology by fabricating carbon nanotube transistors within a conventional wafer fabrication infrastructure.”
IBM has come up with a way to allow more precise and controlled placement of the carbon nanotubes and has now tested a chip with more than 10,000 nanotube transistors on a single chip. Now, that’s still not enough to compete with a modern CPU, but it’s a couple orders of magnitude greater than other experiments have shown.
The important thing about IBM’s new method though is the fact that it can be easily implemented involving common materials and existing manufacturing processes. This could very well drag the idea of a carbon nanotube processor out of the lab and into the commercial chip fabs.